Thanks to advancements in ML, bots are becoming the main interface between humans and machines. 25% of installed apps are never used. The average Android app loses 77% of its daily active users within the first 3 days. The amount of time people spend on social media is constantly increasing. Certainly, the next user interface won't be based on fancy design, it will be based on simple conversation.
4. Interesting Statistics
● 25% of installed apps are never used.
● The average Android app loses 77% of its daily active users
within the first 3 days.
- Google's consumer marketing insight
7. More Interesting Statistics
● Just Yesterday, after we kicked off the conference, Facebook announced that
Messenger now has 1.3 billion active monthly users
● Whatsapp has 1.3 billion active monthly users as well
● 30% of all time spent online is now allocated to social media interaction
● By 2020, over 80% of businesses are expected to have some sort of chatbot
automation implemented (Business Insider, 2016)
19. A. what to do before you start coding
● Understand precisely the role of the bot, make a list of service
● Read the documentation of the platform
● understand what is possible, what is not
24. 1. Break your idea into services
A service is a broad classification of the user requests
your bot would be able to handle. For an e-commerce
company, services could include commerce service,
user complaint service, FAQs service, etc
25. 2. Guidelines for writing services
● A service should be simple enough to describe in a
sentence.
● A service must not be ambiguous, it should have one
and only one meaning
26. 3. Describe each service using intents
Services can further be broken down using intents to
zero-in on exactly what the user wants.
Matching the right intent is very important as this
determines what data to expect/request from the user
and when the conversation can be marked as complete
27. 4. Train a classifier to recognize intents
and services
● Get quality data set for training
● Learn from user interaction
28. 5. Extract Entities and Parameters
● This involves going from sentences to structured data that can serve as
inputs to other systems or microservices.
● Use natural language processing to get names, addresses, time and other
entities of interest.
● Take advantage of suggested responses
● Use slot-filling to handle incomplete request from the user
29. 6. Build a webhook for fulfilment
● After a complete user interaction, the user request has to be processed.
● This typically involves setting up a Rest API
● The API handles fulfilments such as saving the extracted data to the
database, feeding the data to other systems as well as triggering other
independent services, or even a physical action such as a man delivering
pizza.
33. CONCLUSION
Understanding what’s possible and what isn’t with today’s AI and machine
learning capabilities is key for anyone looking to use such technology to build
or enhance applications. If you’re not skeptical enough about claims being
made about current capabilities, you could waste enormous amounts of time
and money trying to do something that can’t (yet) be done. On the other
hand, if you’re too skeptical, you risk missing opportunities to deploy today’s
AI tech in incredibly useful and profitable ways
Source - TechCrunch